Artificial Intelligence from NORCE gives benefits to Norwegian Customs

– We are approaching 9 million customs declarations annually, but some flows are larger. For instance, we receive information about 48 million parcels and letters per year," says Guillaume Lanquepin, director of expertise at Norwegian Customs.

The Norwegian Customs receive and handle all declarations for goods crossing the Norwegian border, both from private and professional purchases and/or sales. Their primary mission is to ensure compliance with laws and regulations for cross-border trade. Their main tasks include verifying compliance with regulations regarding the import and export of goods, ensuring accurate declarations of goods, determining customs duties, and supervising regulations on behalf of between 20 to 30 different public agencies at the border.

And as Lanquepin's figures indicate:

– This results in a large amount of data that must be processed and stored for control and statistical purposes. The data is further used for analysis to detect imports and exports that contravene regulations.

– We are pioneers in expert systems, a branch of artificial intelligence, and already introduced our customs clearance system, TVINN, in the late 1980s. We still actively use AI with image recognition at the border, for example. But we want to better support customs officers, and the current development of AI can contribute to creating better knowledge about risks, Lanquepin explains.

Customs declarations are digitally submitted by employees of companies sending goods to Norway. The manual process of filling out declarations often leads to typos or inaccuracies in the way company names and addresses are entered. This results in many variants of similarly named sending entities, which appear as different entities but potentially represent the same sender in reality.

This makes it challenging to obtain a comprehensive overview of goods imported to and exported from Norway by the same entity. Manual validation of millions of records is not feasible, and this is where artificial intelligence comes into play.

Manual completion of these declarations increases the likelihood of errors due to potential typos or inaccuracies from various sources. As a result, maintaining an accurate and updated archive of foreign entities and their associated declarations presents a significant challenge. The large quantity and frequency of incoming declarations, along with extensive storage requirements, make manual validation or basic fuzzy matching impractical, and this is where artificial intelligence and the project with NORCE come into the picture.

– Machine learning, artificial intelligence, decision support systems, and data analysis are key themes in the project we call "Artificial Intelligence Improves the Customs Authorities' Data Foundation". For us here at NORCE, who have been working with data analysis, machine learning, and decision support systems for decades, collaboration with Norwegian Customs is an enjoyable and challenging task, says Jeremy Cook, senior researcher at NORCE.

Cook further explains:

– This is an Entity Resolution Case project. That is: a branch of artificial intelligence aimed at identifying data reports belonging to the same real-world object.

– Its applications span various areas, including linking customer records, health information, financial transactions, and in this context, identifying foreign senders and recipients. This is especially valuable when combining data from different sources and contributes to customer management and fraud detection. The Customs Authorities benefit from entity resolution by detecting similar foreign entities, thus achieving a more comprehensive data representation of the flow of goods.

Why Are We Doing This for the Customs Authorities?

–This is done to ensure that the goods declarations recorded are correct. For the Customs Authorities, which annually declare large amounts of goods in and out of the country, this is of great importance. Our machine learning algorithms attempt to correct the inaccuracies that occur when goods declarations are filled out and read. The algorithms correct the inaccuracies and/or errors found in already completed declarations," says Cook.

– It is crucial for Norwegian Customs to be able to identify all shipments received or sent by the same actor. This applies to both Norwegian and foreign actors. NORCE supported the Customs Authorities in developing new methods. Entity Resolution is fundamental for integrating information from multiple sources. Linking objects across data sources is not a trivial task. It has also not experienced the same development as generative AI recently. An effective algorithm for Entity Resolution means that the Customs Authorities can better utilize all their data sources, says Lanquepin.

The project with the Customs Authorities is a case study - pilot under EuroCC's national competence centers (NCC), which represent the only contact point at the national level for technology transfer in High Performance Computing (HPC), High Performance Data Analysis (HPDA), and artificial intelligence.

For more information about EuroCC/NCC or to inquire about whether your company qualifies for expert assistance in HPC or AI, please contact the project leader at NORCE, Klaus Johannsen.